1

Growing together or growing apart?A village level study of the impact of the Doha round on rural China

Marijke Kuiper & Frank van TongerenAgricultural Economics Research Institute (LEI) – Wageningen UR

Paper prepared for the 7th Annual Conference on Global Economic Analysis, Trade, Poverty and the Environment, Washington D.C. June17-19, 2004

First draft, May 2004

AbstractMost studies of the opening of the Chinese economy focus on the national level. The fewexisting more disaggregated analyses are limited to analyzing changes in agriculturalproduction. In this paper we employ an innovative village equilibrium model which accountsfor nonseparability of household production and consumption decisions. This allows us toanalyze the impact of trade liberalization on household production, consumption and off-farmemployment, the interactions among these three aspects of household decisions, as well as theinteractions among households in a village economy. Analyzing the impact of tradeliberalization we find changes in relative prices of inputs and outputs and outside villageemployment to have opposite impacts on household decisions. Price changes affect the laborintensity of rice production as households switch between intensive two-season and on-seasonrice in response to changes in their shadow wages. Outside village employment opportunitiesinduce a less labor intensive rice cropping. The consequent drop in demand for tractionservices reduces cash income for those households that do not have access to migration andthus economic growth in coastal areas affects them indirectly through the village factormarket. In terms of aggregate village agricultural supply response we find the impact of theincrease in employment to dominate aggregate response. As a result the marketed surplus ofcase study village is found to be shifting from rice to livestock.

A gradual integration in the global economy, combined with far-reaching domestic reforms, has ledChina to be a showcase of attaining rapid economic growth through market reforms. The rapideconomic growth during the past decades, however, has been accompanied by an increasing disparitybetween coastal and interior regions and between rural and urban areas. The coastal cities benefitedmost from the increasing export opportunities, due to a combination of geographical factors anddeliberate policies (Démurger et al., 2002).

The extent of reform in combination with the size of the Chinese economy, has resulted in a body ofliterature with a growth rate rivaling China’s GDP growth. Most studies of the opening of the Chinese

2

economy to the rest of the world focus at the national level. One exception is a study by Diao et al.(2003) on the regional impact of China’s recent WTO accession. As in other national level studies,they find a positive aggregate effect for China as a whole. This aggregate effect obscures differencesacross regions. Reflecting past trends of diverging growth between coastal and inland provinces (seefor example Démurger et al., 2002; Jones et al., 2003), income gaps between provinces are found towiden following WTO accession. They also find rural-urban migration providing an importantmechanism for transmitting urban growth to the rural areas. Especially central provinces bordering thebooming coastal provinces (Anhui, Jiangxi, Hubei and Hunan), benefit from an increase in rural-urban migration.

A second exception to the common use of a national level analysis is a study by Huang et al. (2003)analyzing the impact of WTO accession on farm households. They find that despite a small positiveaggregate impact of the WTO, the distribution of benefits gives cause for concern. Households inricher coastal areas benefit most, having higher yielding lands and cultivating internationallycompetitive crops. Another interesting finding of their household level analysis is an increase inagricultural production, despite a decrease in the agricultural price index. Farmers respond to thechanges in relative prices induced by the WTO by shifting to more competitive activities (livestock,fish, vegetables and rice), resulting in a net increase in agricultural production.

These two studies show that a disaggregated analysis of the impact of trade reform yields insights thatcannot be attained from national level studies. Our objective in this paper is to contribute a villagelevel perspective on trade reform in China. Adjustment responses of different household types andwithin-village interactions are central to this analysis. This study complements earlier studies byzooming in on the differential impact of trade liberalization on households within a village. A newmethodology that accounts for family-farm production specifics in a village-economy setting gives usa unique perspective on the impact of trade reform. Going beyond the household level study of Huanget al. (2003) we account in addition for the impact of farm income on rural household consumption.We furthermore account for the impact of rural-urban migration, thus not limiting the analysis todevelopments in the agricultural sector.

In this study we combine a macro level analysis of trade reform, with a village level generalequilibrium model of a rice-producing village in Jiangxi Province. To be able to combine the macroand village level analysis of trade reform, we need to ascertain how trade reform affects households.Prices and labor demand form the key transmission mechanisms through which macro level tradereform affects rural households (Winters, 2002). We therefore focus on changes in prices foragricultural inputs and outputs, and on the increased demand for labor by labor-intensive sectors inwhich China has a comparative advantage.

Our analyses of further trade reform builds on a baseline that encompasses China’s recent WTOaccession, phasing out of export quota for textiles under the Agreement on Textiles and Clothing. Weuse the GTAP model to generate macro-level impacts of further reform under a WTO –Doha roundand feed these into the village-level model.

The village equilibrium model incorporates farm-household models, allowing us to analyze the impactof changes in farm production, off-farm employment and household consumption, as well as theinteractions among these three components of household decisions. Since households differ in their

3

endowments and access to off-farm employment, household response to changes in the socio-economic environment differ.

Analyzing the impact of changing in agricultural input and output prices we find an increased villagesupply of rice and other livestock, which corresponds with the findings of national level studies andwith the results of the household analysis in Huang et al. (2003). This increased supply of rice,however, is the net result of two household groups increasing rice production, and two householdgroups reducing rice production. Analyzing the impact of an increase in outside village employmentfollowing trade liberalization, we find it matters whether economic growth is limited to the coastalareas or spread more evenly throughout the country because of differential access to outside provincemigration. The net impact of an evenly spreads increase in employment on supply response is adecrease in rice supply, caused by an increasing scarcity of labor. Again we find diverging householdresponse, with some households increasing rice production due to lower costs of animal tractionrented within the village. The combined impact of price and employment changes following tradeliberalization is a decrease in rice supply from the village. In terms of the village supply response, theimpact of the increased employment thus dominates the impact of the change in prices.

When analyzing the income effects of trade liberalization we find it matters considerably whether weaccount for reduced household consumption due to migrants. Ignoring the consumption of migrants,trade liberalization would reduce within-village income inequality; accounting for the consumptionimpact we find that trade liberalization worsens within-village income distribution.

This paper is structured as follows. We start by providing some background on non-separability ofhousehold decisions and how this affects the structure of the village equilibrium model. The secondsection discusses the data and models used in this study. The third section focuses on the impact ofprice changes, employment and the combination of these two aspects of trade liberalization onhousehold agricultural production decisions. Section four then analyzes the aggregate village supplyresponse as well as the income and welfare impact of trade liberalization. Section five concludes.

1 A household perspective on general equilibrium modeling

As illustrated above, a disaggregated perspective can lead to new insights in the impact of tradereform. Analyzing the household level impact of trade reform, the study of Huang et al. (2003) comesclosest to the approach taken in this study. Due to data limitations their analysis is limited toanalyzing changes in agricultural production, ignoring both changes in household consumptiondecisions and off-farm employment. Separating farm production decisions from householdconsumption decisions is standard in general equilibrium models, both macro as well as existingvillage general equilibrium models. Ignoring the interdependency of household production andconsumption decisions, however, can be misleading when market imperfections render these twoaspects of household decisions nonseparable.

Nonseparability of household production and consumption decisions occurs when the effective priceof a commodity used both in production and consumption is not exogenous to the household, but isdetermined endogenously by household demand and supply. In this case decisions at the production

4

side will affect demand and supply of the commodity, which affects consumption decisions, and viceversa. Such nonseparability occurs if households are not price-takers in a market, if markets aremissing, or if there is a gap between buying and selling prices (Löfgren and Robinson, 1999). Theseminal work of de Janvry, Fafchamps and Sadoulet (1991) shows how rational behavior of farmersin combination with market failures may give rise to sluggish or counterintuitive household responses.

To illustrate a number of points regarding nonseparable household models a simple model ispostulated. The household is assumed to maximize utility derived from the consumption of anagricultural commodity Ca (e.g. a food crop produced by the household) and leisure Cl. The householdfaces a production technology producing the food crop (Qa) with family labor (Lf), hired labor (Lh),and a fixed input ( q ). The household may also work off-farm (Lo), but total labor use cannot exceed

the family time-endowment (T ). The cash constraint provides a third constraint on utilitymaximization; goods can only be bought if money is earned. In this simple model without savings andtransfers this constraint translates to requiring the inflow of money to equal to the outflow.

This very basic household model can mathematically be described as:

),( max,,, laLLCC

CCuhfla

utility function (1)

subject to,

),,( qLLfQ hfa = production function (2a)

TCLL lof =++ time constraint (2b)

haoa wLpCwLpQ +=+ cash constraint (2c)

where, in addition to the variables defined before, p denotes the price of the food crop and w the wagerate. The three constraints may be collapsed into a single full income constraint, by substituting theproduction function and time constraint in the cash constraint, and rearranging:

)]([]),,([ lfhahf CLTLwCqLLfp −−−=− full income constraint (2)

Maximization of the utility function (1) subject to the full income constraint (2) yields input demand,output supply and consumption decisions of the household.

In the mathematical description the interaction between production and consumption decisions is notimmediately obvious. Due to its simplicity the model can also be solved graphically, which doesallow a direct view on (non-)separability of household decisions (Taylor and Adelman, 2003). Themathematical description assumes perfect markets for food and labor, which can be purchased andsold at fixed prices p and w. No distinction is made between own produced and purchased food, norbetween family and hired labor. Due to the fixed prices, production and consumption decisions can beseparated, as is illustrated in Figure 1.

The household faces a trade-off between producing food crop and leisure, depicted by the productionpossibility frontier (PPF). The optimal production level with perfect markets (PMP) is at the pointwhere the slope of the PPF equals the price-ratio between labor and food. Optimal consumption withperfect markets (PMC) is at the point where the marginal rate of substitution equals the price ratio.These are standard micro-economic results from producer and consumer theory, reflecting the

5

neoclassical nature of the model. The graphical representation illustrates that first the productionproblem is solved (moving the line representing the price-ratio until it just touches the PPF), afterwhich the solution to the consumption problem is found (finding the highest indifference curve (I) stilltouching the relative price line).

Source: adapted from Taylor and Adelman (2003)

Figure 1: Household production and consumption decisions with perfect markets (pm) and no markets (nm)for food and labor

The opposite extreme from perfect markets is an absence of all markets. In this case prices are nolonger given, but determined by the interplay of production and consumption decisions. In terms ofFigure 1 this implies finding the point (NM) where the indifference curve Inm touches upon the PPF.In this point the household consumes all the output it produces, and the amount of labor it uses equalsthe time-constraint minus leisure. In this case the prices of food and labor are no longer exogenous.Instead they are endogenous shadow prices (ω and ρ) that cannot be observed in the absence of trade

NM

-w/p

-ω/ρ

PPF

pmlCnm

lC

pmaC

nma

nma CQ =

PMP

PMC

Inm

Ipm

Food crop(Qa, Ca)

Leisure (Cl)T

pmaQ

pml

pmh

pm CTLL −+=

6

with the outside world. The graphical representation illustrates the nonseparability: the optimal pointis now found by simultaneously dealing with consumption decisions represented by the indifferencecurve and production decisions represented by the production possibility frontier.

This basic household model and its graphical solution illustrate a number of points that are essential tofarm household modeling. First, standard economic rules for production and consumption remainvalid. Differences with a separate analysis of production and consumption decisions occur because ofendogenous prices, not because of different behavior by the household. This result implies thatstandard approaches to modeling production and consumption decisions may be also followed.Although the analytical framework remains the same, endogenous household prices complicate thedevelopment of an applied model. The household shadow prices are an analytical construct and thuscannot be directly observed. This complicates the estimation of demand and supply functions fornonseparable household models.

Second, household models tend to generate ambiguous result and easily become analyticallyintractable. Ambiguous results may already occur with perfect markets. In the graphical examplegiven above, assume that prices for food increase The price-ratio line will flatten and the point oftangency to the PPF will be at a higher level of food production. The increase in production and pricesincreases income of the household. The household will be able to shift to a higher indifference curveand increases its food consumption. The increase in food consumption may outweigh the increase infood production, depending on the preferences of the household. The food price increase then doesnot lead to an unequivocal increase in marketed surplus. If an analytical solution of the householdmodel can be derived, it will generally be difficult to sign because of counteracting effects on theproduction and consumption sides of the household. In models with multiple missing marketsinterdependencies between variables will prevent derivation of an analytical solution. Problems inderiving an analytical solution and ambiguous results of analytical solutions, if they can be derived,are important reasons for the development of quantitative household models.

Third, accounting for different levels of market integration of households has clear policy relevance.As the graphical solution in Figure 1 illustrates, integration in markets increases the welfare of thehousehold. Apart from the welfare implications for the household, interaction between producer andconsumer decisions may result in unexpected results. For example, if the objective of a food pricepolicy is to increase the marketed surplus, limited market integration of households may yield adverseresults. Household models help in understanding counterintuitive response to policy changes, as wellas identifying possible complementary policies that promote the desired response.

Nonseparability has become an important feature of household models, but is absent in existinggeneral equilibrium models. A study by Löfgren and Robinson (1999) provides a stylized applicationof including farm-household models in a general equilibrium model, but this approach has not yetbeen implemented in an empirical analysis.

Given the recent transformations in China, markets are still developing and imperfections can beexpected to be abound. Studies of factors influencing migration decisions (Hare, 1999; Murphy, 2000;Rozelle et al., 1999a; Rozelle et al., 1999b) and of patterns in inequality (Benjamin and Brandt,1999), refer to imperfect land, labor and credit markets as being relevant in the Chinese context. Such

7

a partial integration in markets may give rise to nonseparability of household decisions, or may create(thin) local markets through which household decisions affect each other.

If households interact with each other in local village markets, and these markets are not integratedwith markets outside of the village, local general equilibrium effects occur. Studies of marketintegration in China find villages to be integrated in markets for major outputs (Huang et al., 2003)and for fertilizer (Qiao et al., 2003). While villages may thus be assumed to be integrated inagricultural input and output markets, integration of factor markets is limited. Labor markets arehighly segmented (Gilbert and Wahl, 2003), resulting in a rural labor surplus (Cook, 1999). Villagelabor markets, however, can be expected to be limited. Due to the collective ownership of land, allhouseholds have access to land. Lacking landless households and generally having similar agriculturalproduction patterns, there is little scope for local labor markets.

While limiting the prospect for local labor markets, collective ownership of land results in village landmarkets. Land rights remain ambiguous, in spite of land tenure reforms that have granted householduser rights for 30 years, and recently were changed to permit inheritance (Huang and Rozelle, 2004).Land is allocated on the basis of demographic criteria, and readjustments occur to adjust for changesin household size, despite formal household user rights. The result is an ambiguous land tenuresituation (Ho, 2001) in which households have an incentive to keep their land cultivated to avoidloosing it during a next readjustment. Households that migrate to urban areas therefore rent their landto other households, to maintain their claim to the land in case they are unable to secure a living in theurban areas. Given the ambiguity of land tenure, land rental markets are restricted to villagers,resulting in local land markets.

Village interactions may also arise through informal credit markets. Government intervention in theformal banking sector remains strong. Regulated interest rates are well-below market clearing levels,while soft loans to state enterprises seize a large share of available funds. Rural households are thusrationed out of credit markets. In the late 1980s, rural cooperative funds developed, targeted at ruralhouseholds. These funds proofed too successful competitors of existing rural credit cooperatives andwere dissolved in 1999 (Park et al., 2003). As a result of the lack of formal credit options, householdhave to rely on loans from friends and relatives, generating local informal credit markets.

Despite the integration in agricultural input and output markets, local markets for land and capital arethus expected to exist, and imperfect labor markets are expected to result in nonseparability ofhousehold production and consumption decisions. In this study we therefore use a village levelgeneral equilibrium model to account for the interactions among households in local markets, whilepaying due attention to nonseparability of household decisions.

Taylor and Adelman (1996) pioneered the use of general equilibrium models at village level. Theirmodel closely follows the structure of macro level models, for example by modeling production atsector level, which misses the impact of nonseparability of household decisions. We take a differentapproach to village modeling, placing differences in household response due to nonseparability at thecenter of the model. This household perspective on general equilibrium modeling results in a differentmodel structure than used in macro level general equilibrium models and in existing village levelmodels. We model production activities as being household-specific. This allows for idiosyncratichousehold response consistent with nonseparability of production and consumption decisions. A

8

second major difference is the nesting structure used for modeling production decisions. For eachactivity the structure is calibrated on the household survey data1. As a result we have householdspecific production functions, capturing differences in household access to inputs2.

By placing households at the center of the model, the village model used in this study is able tocapture differences in production decisions reflecting differences in access to the inputs, interactionsbetween household production and consumption decisions, and interactions among differenthouseholds within a village economy.

2 Models and data

2.1 Macro level analysis

The macro economic analyses are performed using the GTAP model, with some importantmodifications to the version 5 database. Van Tongeren and Huang (2004) describes the changes madeand the baseline simulation which is the reference for the scenarios used in the present study.Specifically, we have updated policy information on China from the original version 5 GTAPdatabase to include recent changes in domestic policies and changes in border policies. In addition wehave adjusted the Chinese input-output table to incorporate microdata information on primary factorshares. We also made adjustments to the data on livestock production and – consumption. Usingmacro-economic information, and incorporating known imminent policy changes, we project thedatabase forward in time to the year 2010 and beyond. The policy changes along the baseline includethe full implementation of China’s WTO accession and the phasing out of export quota under theAgreement on Textiles and Clothing. This baseline is the reference for simulations to assess theeffects of further trade liberalization under the WTO Doha round.

Our experiment reduces multilaterally the support to agriculture grouped under the three classicalpillars, market access, export competition and domestic support, and it reduces border distortions innon-agricultural sectors. The simulation for agriculture follows closely the original modalitiesproposal tabled in 2002 by the European Commission (European Commission, 2002). This proposalcontained a certain asymmetry between the treatment of OECD countries and developing countries.Specifically, we model a reduction of import tariffs by –36% for trade amongst OECD countries andtotal elimination of import barriers against LDCs, while LDCs themselves are not reducing theirimport barriers. In the area of export competition, we reduce the value of subsidies by -45% in OECDcountries only, and eliminate fully the export subsidies on wheat and oilseeds. In the area of domesticsupport, the AMS ceiling is lowered by -55%, but the blue box payments are maintained. Allreductions are made from the final Uruguay Round commitments. For Non-agricultural Marketaccess, we apply a straight –50% reduction of applied rates. It should be noted that this simulation isillustrative for the possible effects of a successful round. Within the spectrum of proposals offeredprior to the ministerial meeting in Cancun, September 2003, the EC proposal is rather conservative.An assessment of the more far-reaching proposals tabled by the CAIRNS group and the US arediscussed in Van Tongeren and van Meijl (2004). 1 This calibration also rejected the commonly assumed separability of factors and intermediate inputs, and this

assumption was therefore dropped in the village equilibrium model.

9

The main findings for China from the Doha simulation can be summarized as follows. Processedfood, labor intensive horticulture and textiles obtain the highest production gain. In these sectors,China has already a comparative advantage, and exports are an important part of total demand. Forprocessed food tariffs cuts are relatively high and exports are important. We observe a modest outputgrowth in rice. For textiles tariff cuts are modest but exports are very important, so export growthleads to growth production. We also observe an increase in livestock output.

While exports are estimated to grow in some products, the further reduction of remaining importbarriers also enables China to import more products. Imports in such commodities as coarse grains,sugar, cotton, dairy and intensive livestock products are expected to rise.

Table 1: Shocks administered to village model (% change with respect to base)

Price changes EmploymentTrade

liberalizationAgricultural outputs One season rice 2 - 2

Two season rice 2 - 2Other crops 2 - 2Cattle 2 - 2Pigs 2 - 2Other livestock 2 - 2

Agricultural inputs Fertilizer 0 - 0Herbicides 0 - 0Pesticides 0 - 0Seed 2 - 2Purchased feed 2 - 2Other inputs 1 - 1

Consumption goods Food 2 - 2Processed food 2 - 2Nonfood 1 - 1Durables 1 - 1Other expenditures 1 - 1

Wages Non-agricultural employment - 3 3Migration, inside province - 3 3Migration, outside province - 3 3

Outside employment Non-agricultural employment - 9 9Migration, inside province - 9 9Migration, outside province - 7 7

It can be concluded that China can gain a lot from a new WTO round, but most from liberalization inmanufacturing. Around 60% of the potential income gains can be attributed to improved marketaccess for manufacturing goods into OECD countries. Labor intensive manufacturing and foodprocessing industries have the potential to absorb labor and generate incomes in activities outside 2 Detailed description of the village equilibrium model and calibration procedures can be obtained from the

10

primary agriculture. Expanding labor intensive industries in rural areas may be part of a ruraldevelopment strategy.

The simulation is performed relative to the year 2010 baseline, and yields amongst other thingschanges in market prices for key commodities in China, and changes in manufacturing labor demand.These results are fed into the village CGE model. Table 1 summarizes the shocks administered to thevillage model.

2.2 Linking the macro results to the village model

To link the macro analysis to the village model, we follow the conceptual framework of Winters(2002), and focus on prices and labor demand. Studies of market integration of three main staplecrops and fertilizer (Huang et al., 2003; Qiao et al., 2003) show integrated regional and nationalmarkets, with village level prices responding to changes at national level. These integrated commoditymarkets allow us to translate relative price changes derived from the macro level analysis to thevillage level. As in Huang et al. (2003) we use a price transmission coefficient of 0.75 to link macronational level price changes to village level price changes, for commodities that are traded outside thevillage. This implies that a 1 percent change in the national price is modeled as resulting a 0.75percent change in village level prices. A less than unitary transmission coefficient reflects existingfrictions in domestic trade in China. These frictions are expected the persist in the future, given thefindings in Poncet (2003), of a disintegration of domestic markets alongside the continuing integrationof China in international markets.

In addition to the transmission of price changes, we analyze the impact of changes in employmentopportunities. Studies of trade liberalization find an expansion of labor-intensive sectors in whichChina has a comparative advantage. We thus need to link an aggregate expansion of employment tovillage level changes in outside village employment. Lacking data to quantify the link betweennational level changes in employment and household decisions, we assume percentages changes inaggregate labor demand to be completely transmitted to the village. Percentages changes in outsidevillage employment are thus set equal to changes in aggregate labor demand. Such a one-on-onerelation between national level employment and off-farm activities of the households in the case studyvillage seems justified given the findings in Diao et al. (2003). Comparing the change in rural-urbanmigration across regions following WTO accession, they find the fastest increase in rural-urbanmigration in the central provinces. The case study village is located in Jiangxi, one of these centralprovinces, and migration outside of the province plays an important role in the village economy,justifying the assumption of a complete transmission of the demand for labor to the village level.

2.3 The case study village

The case-study village has been selected to be representative of rice producing villages in the plainareas of Jiangxi Province, one of the poorer provinces in China. Data on production and consumptionof 168 households were collected for 2000, using standard household questionnaires with questions

authors.

11

on source and destination of commodities added to allow construction of a village SAM. Thesesurveyed households account for about a quarter of the village population, totaling 729 households.

Differences among households are at the center of our village equilibrium model. Four groups ofhouseholds are distinguished, using ownership of draught power (cattle or tractor) and access tooutside province employment as grouping criteria. The resulting groups represent households withdifferential capacity for earning a living from agriculture and from (transitory) migration to coastalcities.

Table 2 presents the activities of each household type in terms of contribution to value-added. For allhousehold types off-farm employment contributes a significant share of income, but there existimportant differences in the nature of non agricultural income sources. Comparing the two householdgroups lacking a link outside the province, we find households lacking draught power oriented moretowards local off-farm employment, with agricultural activities contributing about 60 percent tovalued-added. The households with draught power obtain about 70 percent of value-added fromagriculture, which is similar to the importance of agriculture for the other household group owningdraught power. The household group with access to migration but lacking draught power derivesabout 40 percent of value-added from outside province migration, and only derives about 45 percentof value-added from agricultural activities. Differential access to agricultural and migrationopportunities is thus reflected in the composition of to value-added.

Table 2: Activity composition (percentage of value-added)Link outside province: No link Link

Owning draught power: No Yes No YesVillage

N= 78 100 256 295 729Agriculture One season rice 9.4 10.8 8.2 10.5 9.5

Two season rice 28.5 28.2 18.6 27.1 23.9Other crops 21.2 24.7 18.4 22.4 20.9Cattle - 7.5 - 8.1 4.1Pigs 0.1 0.1 0.2 0.4 0.3Other livestock 0.1 0.1 0.0 0.0 0.0

Village employment Agricultural labor 1.3 2.2 - 0.1 0.4Local business 19.4 13.5 2.4 2.7 5.2

Outside village Outside employment 18.0 12.9 9.1 5.4 8.9Migration Inside province 2.1 - 4.2 0.9 2.3

Outside province - - 38.8 22.3 24.6100 100 100 100 100

Note: ‘-‘ indicates that the household is not involved in this activity

Differences in activities result in differences in income patterns across the four household groups.Table 3 presents income per adult consumer equivalent3 to allow a direct comparison across

3 Adult equivalent instead of per capita consumption is used to account for differences in consumption between

males and females and between age groups. Lacking survey data, conversion factors were taken from detailed

12

households. The first thing to note is that the household group with an outside link and draught powerhas the highest income per adult consumer equivalent. This is in line with the assumptions made whengrouping households in terms of opportunities for earning agricultural and nonagricultural income.The two household groups without an outside link have similar incomes.

Table 3: Sources of household full income (percentage of total income) and total income (yuan)Link outside province: No link Link

Owning draught power: No Yes No YesVillage

Labor shadow wage income 57.4 52.8 52.6 62.1 57.4above shadow wage income 17.5 12.3 23.3 6.0 13.5

Land irrigated land shadow income 12.4 13.3 12.8 11.4 12.1irrigated land above shadow income 4.8 6.5 1.4 4.5 3.7non-irrigated land 7.3 7.9 7.7 6.6 7.2

Capital Cattle - 3.3 - 3.1 1.8Tractor - 1.6 - 0.8 0.6

Transfer within village transfers 0.6 0.3 0.0 0.5 0.3receipts from outside village 0.0 1.9 2.2 5.0 3.3

Income/adult consumer equivalent (1,000 yuan) 4.4 4.3 5.0 5.5

5.0Note: ‘-‘ indicates that the household does not receive income from this source

Specific features of the village SAM and equilibrium model appear in Table 3, in splitting incomefrom labor and irrigated land between shadow income4 and above shadow income. The householdsurvey data reveal imperfect labor and land markets. Households are involved in a variety of off-farmactivities with different wages. These wages are well above the estimated shadow wage, indicatingrestricted access to off-farm employment and suggesting a situation of labor surplus at the local level.This does not come unexpected, given the high population density in China and similar findings in astudy by Bowlus (2003). In the SAM and village equilibrium model the demand-constrained labormarket is accounted for by valuing labor against household-specific shadow wages, estimated usingthe household survey data. In case of off-farm activities, labor then earns a revenue above the shadowwage which is tracked in a separate account of the SAM.

There is trade in irrigated land (paddy fields), with all four household groups renting in land. Dealingwith a village level analysis we would expect the land market to balance. In total, however,households rent in an amount of irrigated land equal to about a third of their contracted land, that isnot supplied by surveyed households. When conducting the survey using an official list obtained fromthe village administration, a number of the randomly selected households were found to havepermanently left the village. These were replaced with households of which at least some members

consumption data of a study in Bangladesh (Zeller et al., 2001). In addition to differences in age and gender,consumer equivalents were corrected for the length of absence of household members due to migration.

4 Nonseparability results in household-specific shadow prices balancing households’ demand and supply thatcannot be observed. We therefore estimated a production function, explaining the total value of householdoutput by labor, land, manure, feed and external inputs. Based on the estimated production function wecomputed for each household in the sample the household-specific shadow prices for household non-tradablesas the marginal value product of each input. Averaging over the households constituting a household grouprenders a shadow price for each household nontradable that differs across household groups. These shadowprices are used in constructing the SAM and in calibrating the village equilibrium model.

13

were still living in the village, resulting in a biased sample. Lacking households that have moved tothe cities, our sample misses the supply of land by households that have moved to the cities. Asdiscussed in Section 1, the combination of collective ownership with household user rights hasresulted in an ambiguous land tenure system. This applies to the case study village, were according tothe survey data, land reallocations are still occurring. Migrated households thus have an incentive inkeeping their land cultivated by renting it to other households.

Given the ambiguous land tenure system, assuming a perfect land market seems a far stretch from thepattern observed in the data. We therefore estimated a shadow price for irrigated land, which onaverage was about 1.5 times the rent paid for the land. This result fits a distorted land market, whereinsecure land tenure prevents households leaving the village from obtaining a rent reflecting the trueproduction value of the land. At the demand side, this implies that households can rent land for a pricewell below the productive value of the land. In the SAM and village equilibrium model this isaccounted for by having all household groups earning a profit on the land rented in, which is equal tothe difference between the estimated valued of the land and the rent paid. The amount of rented landis kept fixed in the village model, since additional supply is assumed to be associated with themigration of a complete household, which is beyond the scope of this study.

A last remark on the village SAM is the lack of data for modeling capital flows in the village. TheSAM shows that the household group most involved in migration is a net supplier of money for theother three groups of households. Although the survey contains some data on the conditions on whichmoney is provided, insufficient information is available to model a village capital market. In theanalysis money flows among households are therefore assumed fixed at the level observed in theSAM. This excludes one way through which income from migration could be transmitted throughvillage linkages, thus potentially underestimating the village-wide impact of migration.

To summarize the discussion so far, we are analyzing the response of four different types ofhouseholds, distinguished on the basis of their access to agricultural income and income from outsideprovince migration. Analyzing sources of income pointed to imperfect labor and land markets. Theseare accommodated by estimating household-specific shadow prices, introducing profits earned on off-farm employment and renting of land, and by modeling household production and consumptiondecisions as nonseparable.

2.4 The village equilibrium model

Despite introducing nonseparability of household production and consumption decisions, themathematical structure of the model closely resembles macro level general equilibrium models.Consumption decisions are modeled through a linear expenditure system, while production is modeledby nested CES functions. Because of the level of detail in the village model, we do not observe two-way flows of commodities with the outside world. Households consume farm output, but do notpurchase these goods from outside the village, nor from other households in the village. Householdsales to outside village markets are thus equal to total production, minus household consumption.

Off-farm employment options were found to be restricted, resulting in wages exceeding the shadowprice of labor. This is handled in the village equilibrium model by fixing the levels of outside villageemployment and having households earn a profit above labor costs on off-farm activities. Levels of

14

village employment (agricultural and nonagricultural) cannot be fixed, although for these activitieswages also exceed shadow wages. Demand for agricultural labor is therefore assumed to be demand-driven, with prices being fixed5. Demand for village non-agricultural labor is linked to local businessactivities, discussed below.

Village markets exist for traction by draught animals or tractors, and for locally producedconsumption goods. Of these village markets, only animal traction has an endogenous village price inthe model. The SAM indicates that only limited use is made of the tractors. This under-utilization ofavailable tractors is therefore modeled through fixed prices for tractor services, and tractorendowments adjusting to demand (‘unemployment’ closure).

Due to lack of data on other inputs, local business activities use only labor (village non-agriculturallabor), yielding a return that exceeds the shadow wages. All households are involved in local businessactivities and all purchase locally produced goods. This reflects a heterogeneity in goods not capturedby the aggregates used in the SAM and village model. Because of lack of data we fix village prices oflocal goods to deal with the gap between product prices and costs of labor. Assuming fixed pricesseems justified since prices of village produced goods are common knowledge, while shadow wagescannot be observed. Given the unobservable character of shadow wages, it seems unlikely that achange in labor costs will be reflected by a changed village price. A second reason for fixing prices oflocal business activities is the absence of a peak season. Production can therefore be shifted to timeswhen little labor is needed in agriculture, limiting the need to increase the price when shadow wagesincrease.

For all demand-driven activities (local consumption goods, hired agricultural labor, tractor services),market equilibrium is established by allocating demand to suppliers based on the market sharesobserved in the SAM.

Summarizing, the village equilibrium model resembles macro level general equilibrium models by theway in which consumption and production are modeled. A major difference with macro and existingvillage level models is the household-specific production, which is affected by householdconsumption decisions through endogenous household shadow. Lack of data resulted in most villagemarkets to be modeled as demand-driven. The only exception is the village market for animal traction,which is balanced through an endogenous village price. Household production and consumptiondecisions are calibrated on the household survey data, resulting in household-specific demand andsupply functions.

3 Household production response to trade liberalization

Following the analytical framework of Winters (2002) we analyze two pathways through which tradeaffects households: prices and employment. We first separately analyze the impact on householdresponse these two aspects of trade liberalization, followed by an assessment of their combined effect.

5 Agricultural wages are a multitude of shadow wages. Households are therefore always willing to supply

agricultural labor.

15

3.1 Household level impact of price changes

Trade liberalization increases prices of all agricultural outputs and some inputs. In terms of cropproduction, all household groups show a limited response in other crop production, and a significantchange in rice production (Table 4). This difference in response is due to the fixed amount of landused in other crops. This is the only activity which uses non-irrigated land, which is a householdnontradable. This fixed input limits the adjustment possibilities, although the change in shadow priceof non-irrigated land (Table 5) indicates that households are adjusting to the price changes, despitetheir limited production response. This provides a classic illustration of the point made in de Janvry etal. (1991) that, although market imperfections may cause sluggish production response, householdsaccommodate price changes by making internal adjustments.

Table 4: Household production response to price shocks (% change compared to base)Link outside province: No link Link

Owning draught power: No Yes No YesCrops One season rice 3 84 -12 -22

Two season rice 0 -35 6 11Other crops 0 -1 0 0

Livestock Pig production -11 -84 -1 -7Other livestock 148 2420 16 104

In the case of rice, land can be allocated between two different types of rice production. One seasonrice consists of a single crop, while two season rice is a double-cropping system. Consisting of twocrops, the latter uses more inputs, for example for transplanting which is a rather labor intensiveactivity. This difference in labor requirements accounts for the difference between households with nolink outside the province (moving to one season rice) and those with a link outside the province(moving to two season rice).The exogenous increase in rice prices raises shadow wages, which arebased on the value of their marginal product. Ceteris paribus, the increase in the price of rice makestwo season rice more attractive than one season rice, since it has a higher yield. This higher yield,however, is obtained through a more intensive production. Expanding two season rice thus requiresmore labor, increasing shadow wages. For the two household groups lacking an outside link, thisincrease in production costs counterweighs the exogenous increase in output prices.

Table 5: Household-specific price response to price shocks for traded commodities (% change compared to base)Link outside province: No link Link

Owning draught power: No Yes No YesAgricultural output Two season rice 2.0 2.2 2.0 2.0

Pig production 2.0 4.6 2.0 2.0Feed 3.4 9.4 2.4 3.4

Endowments Family labor 2.7 7.2 1.2 4.3Irrigated land 2.6 5.0 2.0 2.7(shadow wages and

rental rates) Non-irrigated land 1.0 -2.4 2.4 0.2

16

As the more moderate shadow wage increases in Table 5 suggests, the two household groups with anoutside link have a larger labor endowment, even after accounting for the absence of householdmembers due to migration. As a result their shadow wage increases less, despite shifting to moreintensive rice cultivation. The shadow wage increase of 7.2 percent for the household with no outsidelink but owning draught power, reflects that this household type has the smallest labor endowment,explaining its shift to one season rice.

If we then turn to the livestock activities, we find pig production to be reduced, while other livestockproduction increases. These different responses have to do with differences in production technology.Pig production is largely based on farm-produced feed, mainly consisting of crop-residues that arerelatively fixed in supply. An increase in pig production will raise the price of feed, being a householdnontradable, making it a less attractive activity. Other livestock uses mainly purchased feed and alimited amount of labor. The feed is purchased outside the village against a fixed price, and otherlivestock production can thus be expanded as long as cash and labor are available.

The household owing draught power and lacking an outside link, shows a dramatic expansion in otherlivestock production. This expansion is less dramatic than it seems at first sight because of the smallsize of this activity in the SAM, accounting for only 0.1 percent of value added. The attraction ofother livestock for this household is that it uses much more purchased feed, as opposed to farm-produced feed, than the other household types. As a result the response of this household is lessaffected by the increasing costs of farm-produced feed, resulting in its strong production response.

Households consume part of their own production. With the linear expenditure system, consumptionof farm output is inelastic. The change in production is thus not matched by a change in consumption,resulting in a stronger change in marketed surplus than in produced output (Table 6). The strongproduction response of the household owning draught power and lacking an outside link, is reflectedby its withdrawal from the market for two season rice and pigs. For these two activities this householdlimits production to its own consumption needs. Not being active in the market, the market price is nolonger relevant. Instead two season rice and pigs have a household-specific shadow price whichexceeds the market price that still holds for the other three household groups (see Table 5).

Table 6: Household marketed surplus response to price shocks (% change compared to base)

The discussion of price changes focused on the impact of changes in agricultural prices. The scenarioanalyzed also includes an increase in the price of consumption goods. Given nonseparability, this

Link outside province: No link LinkOwning draught power: No Yes No Yes

Crops One season rice 12 119 -24 -43Two season rice -4 -100 19 39Other crops -9 -12 1 -5

Livestock Pig production -11 -100 -2 -10Other livestock 332 5123 62 284

17

price increase may affect household production decisions. The impact of the price of consumer goodson production are found to be minor, and are therefore not discussed further.

3.2 Household level impact of changes in outside village employment

We now turn to analyzing the impact of labor demand, providing the second pathway through whichtrade liberalization affects households. Demand for agricultural labor is determined endogenously byagricultural production decisions, and we therefore focus this section on non-agricultural employment.As discussed in the introduction, economic growth in China has concentrated in the coastal provinces,resulting in an uneven distribution of the benefits of growth. Lacking information to link nationallevel changes to regional changes in non-agricultural sectors, we can only make a crude assessment ofthe impact of uneven economic development, through a two-step analysis of changes in non-agricultural employment. To capture the concentration of economic growth in the coastal cities, westart by analyzing the impact of an increase in outside province employment.

Only two of the four households groups have access to outside province employment. Comparing theproduction response between the household groups with and without access to outside province, wefind opposite production responses, which is most pronounced for rice (Table 7). Households thathave an increased opportunity for outside province migration, move from one season to two seasonrice, reducing demand for labor which has become more scarce. The households lacking access tooutside province migration move in the opposite direction by increasing two season rice production.Since these household are not directly affected by the increased migration opportunities, interactionsin local markets cause their response. It turns out that interactions in the village market for animaltraction transmit the effects of the increased migration, and therefore mainly affect the householdowning draught power. The shift to one season rice by the migration households reduces the demandfor animal traction, resulting in a decrease of the price for traction. The two households not involvedin migration do not face a increased shadow wage, which combined with the lower costs for tractionmakes the more intensive two season rice production more attractive.

Table 7: Household production response to increase in outside province migration (% change)Link outside province: No link Link

Owning draught power: No Yes No YesCrops One season rice -2 -22 5 77

Two season rice 1 10 -2 -34Other crops 0 0 -1 0

Livestock Pig production 0 0 3 -2Other livestock -5 -47 -24 93

The reduction in the price of animal traction also reduces cash income of the two household groupsowning draught power. For the household group with access to migration, this decrease in cashincome is outweighed by an increase in remittances. A net increase in cash remains, which incombination with the higher shadow wage results in an expansion of other livestock production. Thehousehold lacking access to migration faces a reduction in the available cash due to the decreased

18

incomes from renting animal traction services. This household group also has the most capitalintensive other livestock production, resulting in a contraction of this activity.

Only shocking outside province employment thus shows that because of asymmetric access to outsideprovince migration and village interactions, opposing production changes occur. Despite aconcentration of economic growth in coastal cities, other regions benefit from trade liberalization aswell. Since all household types are involved in off-farm employment, they will all benefit from anincreased demand for nonagricultural labor outside the village. An increase in all types of off-farmemployment, either inside or outside the province, has a similar impact as found above for the twohouseholds with an outside province link (Table 8). Migration outside the province accounts for about70 percent of their off-farm employment income, and the analysis above about covers the impact of anincrease in outside village employment.

Table 8: Household production response to increase in outside village employment (% change compared to base)Link outside province: No link Link

Owning draught power: No Yes No YesCrops One season rice -15 15 8 77

Two season rice 5 -6 -4 -33Other crops 0 0 -1 0

Livestock Pig production 2 3 4 -1Other livestock -44 -47 -27 78

Of the two households lacking an outside province link, the household owning draught animals showsa qualitatively different response in terms of rice production, than when analyzing outside provincemigration. This household type has a limited involvement in outside village employment (accountingfor 12 percent of its value-added), but having the smallest labor endowment its shadow wage doesincrease when off-farm employment increases. At the same time, the demand for animal tractiondrops further, due to the twice as strong reduction in two-season rice of the household with an outsidelink but not owning draught power. An increasing shadow wage combined with a reduction in cashincome that is only partly compensated by an increased off-farm income, the household owningdraught power but not having access to migration increases one season rice production. This shiftreduces demand for labor and cash, that have both become scarcer.

Summarizing, an increase in outside village employment has a differential impact on householdsgroups, depending on their factor endowments and access to different types of off-farm employment.Separately analyzing an increase in migration to coastal provinces shows that the access toemployment by two household groups, triggers a strong production response of the householdslacking such access. Local markets for animal traction provide the main link through which thesechanges are transmitted through the village economy. If the increased demand for labor is notconcentrated in the coastal cities, all households benefit from an increase in off-farm income.Concentrated versus evenly distributed increase in labor demand qualitatively changes the response ofone household type, pointing to a need for a more disaggregated analysis of the impact of tradeliberalization.

19

3.3 Household level impact of trade liberalization

Analyzing the two pathways through which trade liberalization affects households resulted inopposing production responses (compare Table 4 and 8). This raises the question what the net impactof these two types of shocks are. Major changes occurred with rice and livestock production, for theseactivities we therefore indicated whether the increase in agricultural prices or the increase in outsidevillage employment dominates household response. From the entries in Table 9 we can conclude thatthe dominant aspect of trade liberalization depends on the household type.

Table 9: Household production response to trade liberalization (% change compared to base) (*)Link outside province: No link Link

Owning draught power: No Yes No YesCrops One season rice -32 (o) 83 (i) -5 (a) 78 (o)

Two season rice 11 (o) -34 (i) 4 (a) -33 (o)Other crops -1 -1 -1 -0

Livestock Pig production -7 (a) -84 (i) 3 (o) -9 (i)Other livestock 84 (a) 2398 (i) -15 (o) 209 (i)

Note: (a) means the change in agricultural prices dominates; (o) means the change in outside village employment dominates;(i) means prices and employment have an identical impact in qualitative terms. (*) combined effect of price changes andoutside village employment.

In case of the household type owning draught power and lacking a link outside the province, the twoaspects of trade liberalization have the same qualitative impact. The total impact is almost identical tothe impact of changing agricultural prices, which is due to the limited involvement in off-farmemployment of this household type. The other household group owning draught power has a similarpattern of shifts in agricultural production. In this case the impact of outside village employmentdominates, which is not surprising since this already accounts for a quarter of its value-added in thebase run. Both household groups owning draught power thus shift to one season rice and otherlivestock production, but this is because of change in agricultural prices for one household type whilebeing due to changes in outside province migration for the other.

The two household groups not owning draught power display an opposite shift in crop production, bymoving to two season rice. Both earn a significant share of their value-added from outside villageemployment. Surprisingly, for the household involved in migration the impact of the increase inemployment (accounting for 50 percent of value-added) does not dominate its response. The increasein rice prices combined with the decreased costs of animal traction outweighs its increased shadowwage. The importance of off-farm employment thus does not fully determine which aspect of tradeliberalization dominates.

The other household type that does not own draught power earns a smaller part of its income throughoff-farm employment, and therefore has a more modest increase in its shadow wage (Table 10). Forthis household type the decreased price of animal traction following an increase in employmentoutweighs the rise in labor costs, resulting in a shift to two season rice.

20

Table 10: Household-specific price response to price shocks for traded commodities (% change compared to base)

Link outside province: No link LinkOwning draught power: No Yes No Yes

Agricultural output Two season rice 2.0 2.2 2.0 2.1Pig production 2.0 4.2 2.0 2.0

Endowments Family labor 8.8 10.3 14.3 8.9Irrigated land 1.7 5.2 2.6 3.7(shadow wages and

rental rates) Non-irrigated land -0.8 -3.9 -2.3 -0.2Animal traction services n.a. -2.6 n.a. -2.6

Three of the households groups increase other livestock production. This activity uses a lot of externalinputs and is constraint by the availability of cash needed for purchasing external inputs. The increasein output prices and outside village employment release this cash constraint, resulting in anspectacular percentage increase in other livestock production. Actual marketable amounts are moremodest since it only accounts for a fraction of total output in the base run (see below). The householdwith an outside link and not owning draught power increases pig production while reducing otherlivestock production. This different response is due to the household-specific calibration of theproduction functions. As a result this household type gains more from increasing the amount ofpurchased feed in pig production than in other livestock production.

4 Village level impact of trade liberalization: supply response and income distribution

The previous section focused on differences across household types. This discussion illustrated thatdifferences in household endowment and access to off-farm employment affect household productionresponse. Finding opposing production responses to agricultural price changes and employmentchanges, begs the question what the aggregate village supply response to trade liberalization is. Asecond question that comes to mind is which household gains most from the trade liberalization. Thesecond part of this section therefore discusses the income and welfare impacts of trade liberalization.

4.1 Village agricultural supply response

As with analyzing household response, we look at price and employment effects of trade liberalizationseparately before assessing the combined effect on the village marketed surplus (Table 11). We alsocomputed the change in total marketed surplus of rice, summing one and two season rice.

The discussion of household response was cast in terms of percentage changes, ignoring the differentsizes of the household groups. As indicated in Table 2, the two household groups with an outside linkare larger than the two household groups lacking such a link. This difference in size affects thecontribution of each household group to the village marketed surplus. The household type with anoutside link and owning draught power forms the largest group (295 households), and the villagemarketed surplus follows the same pattern as the production response of this household group.

21

Table 11: Village marketed surplus with trade liberalizationPrice changes

(1)Employment changes

(2)Trade liberalization(1) and (2) combined

Village marketed surplus (% change compared to base):One season rice -5 63 73Two season rice 6 -51 -57

Rice 2 -14 -15Other crops -3 -6 -8Pig production -16 2 -13Other livestock 1214 2 1253

Composition of village marketed surplus (% of total surplus):One season rice 13 32 25Two season rice 26 17 11Other crops 9 11 8Pig production 22 37 23Other livestock 30 3 32

100 100 100

The change in agricultural prices following trade liberalization increases total exports of rice from thevillage, due to a shift to two season rice by the two largest household groups. This shift is induced byan increase in rice prices, with most other input prices remaining unaffected. This asymmetric pricechanges makes intensive rice cultivation more attractive. The increase in other livestock production islargely due to the explosion of the marketed surplus by the household owning draught power butlacking an outside province link (see also Table 6), which is only partly tempered by the relativelysmall size of this household group (100 households).

The increased labor demand following trade liberalization has the opposite impact of agriculturalprice changes. The increasing shadow wages result in a shift to less intensive rice cultivation, andtotal rice exports from the village drop by 14 percent. Differences across household types in marketedsurplus of other livestock almost cancel each other, leaving a very modest 2 percent increase.

Combining price and employment changes, the impact of increased employment dominates themarketed surplus of crops, while agricultural price changes dominate the impact on the marketedsurplus of livestock activities. The net result is a decrease in rice exports from the village, whileexports from other crops also drop with 8 percent. Declines in household production response of othercrops were modest for all three scenarios. But after summing over all households, and accounting forincreased household consumption following the increased incomes, the reduction becomes significantat village level.

The explosion in marketed surplus of other livestock in percentage terms raises questions on the sizeof the other livestock sector following trade liberalization. Table 2 indicates a limited share of otherlivestock in production in the village SAM, which amounts to 3 percent of the value of villagemarketed surplus. The explosion of other livestock production results in ten-fold increase in its

22

contribution to the value of village marketed surplus. Simultaneously rice production drops inimportance, accounting for 50 percent of marketed surplus in the SAM, which drops to 36 percentfollowing trade liberalization. In terms of total marketed surplus, the village thus can be characterizedas shifting from rice-producing to a livestock producing village.

4.2 Household income and welfare

The discussion so far has focused on changes in agricultural production which was found to shiftconsiderably in response to changes in prices and employment. The impact of these changes onhousehold income and welfare has been left untouched. Since growing income inequalities are a risingconcern in China, we now turn to analyzing income and welfare. Again, we split the impact of tradeliberalization in prices, before looking at their combined impact.

Table 12: Changes in full income and welfare following trade liberalization per household and village total

Prices(1)

Employment(2)

Trade(1) and (2)combined

Full income per household (% change):No draught power, no outside province link 2.5 4.2 7.0Draught power, no outside province link 4.9 1.9 7.0No draught power, outside province link 1.3 8.0 9.3Draught power, outside province link 3.2 3.7 6.1

Total1 increase in full income (% change): 2.7 4.9 7.4

Welfare increase per household (EV, 1,000 yuan):No draught power, no outside province link 0.3 0.5 0.9Draught power, no outside province link 0.7 0.3 1.0No draught power, outside province link 0.2 1.3 1.5Draught power, outside province link 0.6 0.8 1.2

Total1 welfare increase (EV, 1,000 yuan): 313.9 630.0 910.01 Computed as the sum over all households.

The upper-part of Table 12 presents percentage changes in full income per household. The householdlacking an outside link but owning draught power benefits most from the agricultural price changes.This is a direct result of its activity composition (see Table 2), which focuses on agriculturalproduction. Not surprising given its orientation on outside province migration, the household with anoutside link but lacking draught power benefits least from the price change, but most from theincreased employment opportunities. The total increase in full income is considerable for allhouseholds, ranging from 6 to 9 percent.

The lower part of Table 12 presents the welfare increase per household, measured by the equivalentvariation. The pattern is similar as with the changes in full income. The households owning draughtpower are oriented towards agriculture and benefit most from the price changes, while the householdsinvolved in outside province migration benefit most from the increased employment opportunities.

23

The total increase in welfare following trade liberalization varies between 900 and 1,500 yuan perhousehold.

Table 13: Full income and cash income per adult equivalent following trade liberalization (,000 yuan)Base Prices Employment Trade

Full income per adult equivalent:No draught power, no outside province link 4.3 4.4 4.5 4.6Draught power, no outside province link 4.3 4.5 4.4 4.6No draught power, outside province link 3.8 3.9 4.1 4.2Draught power, outside province link 4.3 4.4 4.4 4.5Village average 4.1 4.2 4.3 4.4

Full income per adult equivalent adjusted for migration:No draught power, no outside province link 4.4 4.5 4.6 4.7Draught power, no outside province link 4.3 4.5 4.4 4.6No draught power, outside province link 5.0 5.0 5.5 5.6Draught power, outside province link 5.5 5.7 5.8 5.9

Cash income per adult equivalent adjusted for migration:No draught power, no outside province link 0.8 0.8 0.9 0.9Draught power, no outside province link 0.8 0.8 0.8 0.8No draught power, outside province link 1.7 1.7 1.9 1.9Draught power, outside province link 1.1 1.2 1.2 1.2

The full income and welfare per household indicate that although all households benefit from thetrade liberalization, welfare gains are concentrated with the households with access to outsideprovince migration. To assess the impact on the income distribution within the village, Table 13presents different measures of income per capita. The upper-part presents the full income per adultequivalent, which is the household size measured in male-equivalent consumer units ignoring theabsence of migrated household members. As can be expected given the egalitarian distribution ofland, income differences are not huge. Note that in terms of income per adult, the household typeinvolved in outside province migration and lacking draught power is the poorest household. Since thishousehold benefits most from the trade liberalization, this would suggest that access to rural-urbanmigration reduces the income differences.

Measuring income, however, is a tricky matter. The middle part of Table 13 presents an alternativeincome measure, adjusting household size for the absence of migrants6This turns the ranking ofhouseholds up-side down and the income distribution appears much less egalitarian. Using hismeasure, households involved in outside province migration are the richer households to start withand income differences grow with trade liberalization. Accounting for reduced consumption demandwith an increase in migration thus increases both the gains from trade liberalization as the widening ofincome differences.

6 The household survey data only include remittances received from migrants, we thus lack data on the actual

income received by the migrants. The remittances are net of the consumption expenditures by the migrantsmade while being away. We can thus deduct the time which migrants are absent from the consumption needsof the households.

24

Through incorporating farm-household models in the village model we are able to measure the fullincome of households, including the consumption of farm output. Measurements of income or povertygenerally need to rely on recorded expenditures, and we therefore included a measure of the cashincome of households in the bottom part of Table 13. The cash income is measured as the totalexpenditures on consumption goods, purchased inside or outside the village. Comparing thesenumbers to the full income measures shows the importance of the consumption of farm producedoutput. Comparing across scenarios, the impact of trade liberalization appears limited when measuredin terms of cash expenditures. Again, because of the increase in migrants, the household with anoutside link and lacking draught power gains most from trade liberalization.

5 Conclusions

In this study we employed an innovative village equilibrium model, which fully accounts for theimpact of nonseparability on household production and consumption response. This allowed us toanalyze the impact of trade liberalization on agricultural supply response and off-farm employment,while accounting for household consumption decisions. To our knowledge this is the first empiricalvillage level study that accounts for nonseparability of household production and consumptiondecisions, both in construction of village SAM, the structure of the village model and the calibrationof household production and utility functions. The village model is used to analyze the impact of tradeliberalization which was quantified through a macro level GTAP model. We analyzed the impact ofprice changes and labor demand, the two major pathways through which international trade affectshouseholds.

Analyzing the impact of changes in agricultural prices we find that labor endowments determine thechoice for more or less intensive rice cultivation. Two household groups shift to less intensive oneseason rice production, since an increase in shadow wages outweighs the increase in rice prices. Theother two household groups move in the opposite direction moving to more intensive rice production.Since the latter two groups are larger, the village marketed surplus of rice increases in response to theagricultural price changes. The village marketed surplus of other livestock also increases, mainly dueto an explosion of this activity by one type of household which uses a lot of external inputs and littlelabor in producing other livestock. The rise in output prices increases the availability of cash, allowingthis household to expand its production.

Analyzing the impact of an increase in employment we find that it matters whether the increaseddemand for labor is concentrated in the coastal cities, or spread more evenly throughout the provinces.Comparing household production response to an increase in outside province employment, we findthat households having access to migration move to less labor intensive rice production, whilehouseholds lacking access to migration move to more labor intensive cultivation. The response of thelatter group of households is triggered by a decreased price of animal traction services, due to the less-intensive cultivation of households involved in migration. If both inside and outside provinceemployment increases, one type of household qualitatively changes it production response. A rise inits shadow wage combined with a further decrease in demand for animal traction, has the householdowning draught power but lacking an outside province link to shift to one season rice as well. The netimpact of an increase in employment is a reduction in the village-marketed surplus of rice.

25

The two pathways through which trade affects households thus have an opposite impact on householdproduction response. Assessing the combined effect at household level, the dominant aspect of tradeliberalization depends on household endowments and on production activities. We furthermore foundthat a strong involvement in off-farm employment does not necessarily imply that the employmentaspect of trade liberalization dominates household response, hampering a ex ante judgment on themost relevant aspect of trade liberalization for a specific household type.

Rising income inequalities are a growing concern in China. Analyzing the income and welfareimpacts of trade liberalization we find that although all households benefit, the households involved inoutside province migration gain most. Whether this improves or worsens within-village incomedistribution depends on whether one accounts for the reduction in consumption demand whenhousehold members migrate. If the consumption impact of migration is ignored, income distributionimproves with trade liberalization. Accounting for the absence of migrants leads to the oppositeconclusion, with the richer households gaining most from trade liberalization. Thus although rural-urban migration can transfer benefits from economic growth in the coastal provinces to inlandprovinces, asymmetric access to migration implies that the rising rural-urban income differences aretransferred as well.

The findings of this study clearly illustrate the advantages of using a village equilibrium model thatallows for household-specific production. Opposite movements in production, reflecting differenthousehold endowments, temper the village level production response. The impact of differentialhousehold endowments would have been missed with a sector level modeling of production whichseparates firm production decisions from household consumption decisions. The analysis alsohighlights the importance of local links in transmitting shocks throughout the village economy,despite the village being well integrated in agricultural output and input markets.

Given the relatively standard data needed for the model, the most important limitation for applyingthis methodology elsewhere is the amount of detail at village level, which comes at the cost ofcoverage. A relevant question is whether the results of the case study village are representative forvillages outside Jiangxi province. An interesting avenue for answering this question it to use themodel as a laboratory for understanding mechanisms and for formulating testable hypotheses on theimpact of trade on household incomes and production decisions. Using data collected in othervillages, these hypotheses could then be tested to see whether the findings from the villageequilibrium model can be generalized.

6 References

Benjamin, D., and Brandt, L. (1999). “Markets and inequality in rural China: parallels with the past”in The American Economic Review, Vol. 89, No. 2: 292-295.

Bowlus, A. J., and Sicular, T. (2003). “Moving Towards Markets? Labor Allocation in Rural China”in Journal of Development Economics, Vol. 71: 561-583.

Cook, S. (1999). “Surplus labour and productivity in Chinese agriculture: evidence from householdsurvey data” in The Journal of Development Studies, Vol. 35, No. 3: 16-44.

26

de Janvry, A., Fafchamps, M., and Sadoulet, E. (1991). “Peasant household behavior with missingmarkets: some paradoxes explained” in Economic Journal, Vol. 101: 1400-1417.

Démurger, S., Sachs, J. D., Woo, W. T., Bao, S., and Chang, G. (2002). “The relative contributions oflocation and preferential policies in China's regional development: being in the right place andhaving the right incentives” in China Economic Review, Vol. 13: 444-465.

Diao, X., Fan, S., and Zhang, X. (2003). “China's WTO accession: impacts on regional agriculturalincome - a multi-region, general equilibrium analysis” in Journal of Comparative Economics,Vol. 31: 332-351.

European Commission. (2002). “The EC's Proposal for Modalities in the WTO AgricultureNegotiations.” Ref. 625/02, Directorate-General for Trade / Directorate-General forAgriculture, Brussels.

Gilbert, J., and Wahl, T. (2003). “Labor market distortions and China's WTO accession package: anapplied general equilibrium assessment” in Journal of Comparative Economics, Vol. 31: 774-794.

Hare, D. (1999). “'Push' versus 'pull' factors in migration outflows and returns: determinants ofmigration status and spell duration among China's rural population” in The Journal ofDevelopment Studies, Vol. 35, No. 3: 45-72.

Ho, P. (2001). “Who Owns China's Land? Policies, Property Rights and Deliberate InstitutionalAmbiguity” in The China Quarterly, Vol., No. 166: 394-420.

Huang, J., Li, N., and Rozelle, S. (2003). “Trade Reform, Household Effects, and Poverty in RuralChina” in American Journal of Agricultural Economics, Vol. 85, No. 5: 1292-1298.

Huang, J., and Rozelle, S. (2004). “Agricultural Development and Policies in China.” in F. W. v.Tongeren and J. Huang (eds.) China's food economy in the early 21st Century; Developmentof China's food economy and its impact on global trade and on the EU. AgriculturalEconomics Research Institute (LEI), Report 6.04.04, The Hague.

Jones, D. C., Li, C., and Owen, A. L. (2003). “Growth and regional inequality in China during thereform area” in China Economic Review, Vol. 14: 186-200.

Löfgren, H., and Robinson, S. (1999). “Nonseparable farm household decisions in a computablegeneral equilibrium model” in American Journal of Agricultural Economics, Vol. 81, No. 3:663-670.

Murphy, R. (2000). “Migration and Inter-household Inequality: Observations from Wanzai County,Jiangxi” in China Quarterly, Vol. 0, No. 164: 965-82.

Park, A., Brandt, L., and Giles, J. (2003). “Competition under credit rationing: theory and evidencefrom rural China” in Journal of Development Economics, Vol. 71, No. 2: 463-495.

Poncet, S. (2003). “Measuring Chinese domestic and international integration” in China EconomicReview, Vol. 14: 1-21.

Qiao, F., Lohmar, B., Huang, J., Rozelle, S., and Zhang, L. (2003). “Producer Benefits from InputMarket and Trade Liberalization: the Case of Fertilizer in China” in American Journal ofAgricultural Economics, Vol. 85, No. 5: 1223-1227.

Rozelle, S., Guo, L., Shen, M., Hughart, A., and Giles, J. (1999a). “Leaving China's farms: surveyresults of new paths and remaining hurdles to rural migration” in The China Quarterly, Vol.,No. 158: 367-393.

Rozelle, S., Taylor, J. E., and deBrauw, A. (1999b). “Migration, remittances, and agriculturalproductivity in China” in The American Economic Review, Vol. 89, No. 2: 287-291.

Taylor, J. E., and Adelman, I. (1996). Village economies: the design estimation and use of villagewideeconomic models, Cambridge University Press, Cambridge, etc.

Taylor, J. E., and Adelman, I. (2003). “Agricultural Household Models: Genesis, Evolution, andExtensions” in Review of Economics of the Household, Vol. 1: 33-58.

van Tongeren, F. W., and Huang, J. (2004). “China's food economy in the early 21st Century;Development of China's food economy and its impact on global trade and on the EU.”,Agricultural Economics Research Institute (LEI), Report 6.04.04, The Hague.

27

van Tongeren, F. W. and H. van Meijl (2004). "Beyond the Uruguay Round: An evaluation of DohaDevelopment Round proposals", in F. W. v. Tongeren and J. Huang (eds.) China's foodeconomy in the early 21st Century; Development of China's food economy and its impact onglobal trade and on the EU. Agricultural Economics Research Institute (LEI), Report 6.04.04,The Hague.

Winters, L. A. (2002). “Trade, Trade Policy and Poverty: What Are the Links?” in The WorldEconomy, Vol. 25: 1339-67.

Zeller, M., Sharma, M., Ahmed, A. U., and Rashid, S. (2001). “Group Based Financial Institutions forthe Rural Poor in Bangladesh. An Institutional and Household Level Analysis.” ResearchReport 120, IFPRI, Washington D.C.